The following may be what you are looking for, although this allows the match and condition key words to be on the same line. If that's not a requirement the code could be simplified somewhat.

#!/usr/bin/perl
use strict;
use warnings;

my %matches = (I => {cond => 'megabyte'}, code => {cond => 'I'},);
my $match = "\\b" . join("\\b|\\b", keys %matches) . "\\b";
my %conds;
my $condMatch;

while (defined(my $line = <DATA>)) {
    my @segments = split /(?=$match)/, $line;

    for my $segment (@segments) {
        while ($segment =~ /($match)/g) {
            my $cond = $matches{$1}{cond};

            if (exists $conds{$cond}) {
                delete $conds{$cond};
            } else {
                $conds{$cond} = $line;
            }

            $condMatch = join "\\b|\\b", keys %conds;
            $condMatch = "\\b$condMatch\\b" if $condMatch;
        }

        if ($condMatch && $segment =~ /($condMatch)/) {
            print $conds{$1};
            delete $conds{$1};
        }
    }
}

__DATA__
Suppose I want to find something, that is present multiple times in a 
+large
(several hundred megabyte) file. However, I only want it if something 
+else
exists after it but before the next occurrence of the thing I am looki
+ng for.
For instance the thing I am looking for might be a set of code numbers
+ and the
condition that I will use to decide whether I want the code may be 1 o
+r more
lines further into the file but before the next occurrence of a code n
+umber.
What is the best (most elegant) way to approach this problem? Chet
[download]

Prints:

Suppose I want to find something, that is present multiple times in a 
+large
For instance the thing I am looking for might be a set of code numbers
+ and the
[download]